Title :
Fixed-point square roots
Author :
Seth, Abhishek ; Gan, Woon-Seng
Author_Institution :
Digital Signal Process. Lab., Nanyang Technol. Univ., Singapore, Singapore
Abstract :
Square root (SQRT) is a common arithmetic operation used in many DSP algorithms. In this paper, we evaluate square rooting methods suitable for implementation on fixed-point (FxP) DSP processors with a fast multiplying unit. The finite wordlength effect on the square rooting methods is highlighted, and it is shown that the theoretically derived convergence rate for the Newton-Raphson (NR) based square rooting methods are not suitable for FxP processor. Also, the most efficient methods for 8-bit and 16-bit FxP processors are identified.
Keywords :
Newton-Raphson method; digital signal processing chips; fixed point arithmetic; DSP algorithms; Newton-Raphson based square rooting method; arithmetic operation; convergence rate; finite wordlength effect; fixed-point DSP processors; fixed-point square roots; Chebyshev approximation; Convergence; Digital signal processing; Program processors; Signal processing algorithms; Table lookup; DSP; Fixed-point; bit precision; square root;
Conference_Titel :
Acoustics, Speech and Signal Processing (ICASSP), 2012 IEEE International Conference on
Conference_Location :
Kyoto
Print_ISBN :
978-1-4673-0045-2
Electronic_ISBN :
1520-6149
DOI :
10.1109/ICASSP.2012.6288231
